用于头颈癌动态准直器旋转VMAT中静态角度调制端口配置的自动化决策支持工具

文献信息

中文摘要

背景： RapidArc Dynamic（RAD；Varian Medical Systems，Palo Alto，CA）是一种新型容积旋转调强放疗（VMAT）技术，其特征是准直器旋转与机架旋转同步。RAD能够使用由用户选择的机架角和准直器角定义的静态角度调制端口（STAMP）。然而，目前STAMP配置的确定依赖于基于计划者经验的手动试错流程，这限制了该功能的高效临床实施。
目的： 本研究旨在开发一种使用Eclipse脚本应用程序编程接口（ESAPI；Varian Medical Systems）的自动化决策支持工具，以识别成本优化的STAMP配置，并评估其在复杂头颈癌计划中的临床实用性。
方法： 回顾性分析20例既往在本机构接受过放疗的鼻咽癌或鼻窦癌患者。为每例患者生成四种治疗计划：使用RapidArc（RA，Varian Medical Systems）的常规VMAT、由专家放射肿瘤科医生手动定义STAMP的RAD（RADm）、使用供应商提供的自动准直器旋转算法的RAD（RADa），以及使用所提出的决策支持工具的RAD（RADdst）。所开发的算法分三个阶段确定STAMP配置：（1）通过整合计划靶区适形度和危及器官（OAR）规避项，生成跨机架角和准直器角的二维几何成本图；（2）提取成本最小化的准直器轨迹；（3）在需要大幅准直器旋转且满足机器特定旋转速度约束的几何关键区域中，识别成本优化的STAMP位置。评估靶区和OAR的剂量学指标、优化计算时间以及估计的束流交付时间。
结果： 基于ESAPI的工具成功为所有患者生成了剂量学可行的RADdst计划，根据解剖复杂性自动选择2至10个STAMP。所有计划技术均实现了临床可接受的靶区覆盖和OAR保护。在鼻咽癌病例中，与RA相比，所有RAD方法均显著降低了口腔剂量（RADm，p = 0.02；RADa，p = 0.002；RADdst，p = 0.002），并且RADdst还实现了左晶状体剂量的显著降低（p = 0.008）。在鼻窦癌病例中，七项OAR剂量学指标在技术间观察到显著差异（p < 0.05），而靶区覆盖在所有技术间保持相当。与RA相比，所有RAD方法的优化计算时间减少了约40%（p < 0.01）。此外，与RA相比，所有RAD技术将估计的束流交付时间减少了约20%（p < 0.01）。
结论： 成功开发了一种基于ESAPI的自动化决策支持工具，用于RAD计划中的STAMP配置。所提出的基于几何成本的框架实现了与专家定义的手动STAMP计划和供应商提供的自动准直器旋转方法相当的剂量学性能，同时消除了手动确定STAMP的需要。这些发现支持自动化STAMP配置作为一种实用策略的可行性，可实现RAD计划的高效且独立于计划者的实施。

英文摘要

BACKGROUND: RapidArc Dynamic (RAD; Varian Medical Systems, Palo Alto, CA) is a novel volumetric modulated arc therapy (VMAT) technique that features dynamic collimator rotation synchronized with gantry rotation. RAD enables the use of static angle-modulated ports (STAMPs) defined by user selected gantry and collimator angles. However, the determination of STAMP configurations currently relies on manual trial-and-error procedures based on planner experience, which limits the efficient clinical implementation of this functionality.
PURPOSE: This study aimed to develop an automated decision support tool using an Eclipse Scripting Application Programming Interface (ESAPI; Varian Medical Systems) to identify cost-optimized STAMP configurations and evaluate its clinical utility in complex head and neck cancer planning.
METHODS: Twenty patients with nasopharyngeal or sinonasal cancer who had previously received radiotherapy at our institution were retrospectively analyzed. Four treatment plans were generated for each patient: conventional VMAT using RapidArc (RA, Varian Medical Systems), RAD with manually defined STAMPs (RADm) by an expert radiation oncologist, RAD with a vendor-provided automatic collimator rotation algorithm (RADa), and RAD using the proposed decision support tool (RADdst). The developed algorithm determined the STAMP configurations in three phases: (1) generation of a two-dimensional geometric cost map across the gantry and collimator angles by integrating planning target volume-fit and organ-at-risk (OAR)-avoidance terms, (2) extraction of a cost-minimizing collimator trajectory, and (3) identification of cost-optimized STAMP positions in geometrically critical regions that require substantial collimator rotation while satisfying machine-specific rotation-speed constraints. The dosimetric indices of the targets and OARs, optimization calculation times, and estimated beam delivery times were evaluated.
RESULTS: The ESAPI-based tool successfully generated dosimetrically feasible RADdst plans for all patients, automatically selecting between two and ten STAMPs according to anatomical complexity. All planning techniques achieved clinically acceptable target coverage and OAR sparing. In nasopharyngeal cancer cases, all RAD approaches significantly reduced the oral cavity dose compared with RA (RADm, p = 0.02; RADa, p = 0.002; RADdst, p = 0.002), and RADdst additionally achieved a significant reduction in the left lens dose (p = 0.008). In sinonasal cancer cases, significant differences were observed among the techniques for seven OAR dosimetric metrics (p < 0.05), whereas the target coverage remained comparable across all techniques. The optimization calculation time was reduced by approximately 40% for all RAD approaches compared with RA (p < 0.01). Furthermore, all RAD techniques reduced the estimated beam delivery time by approximately 20% relative to RA (p < 0.01).
CONCLUSIONS: An automated ESAPI-based decision support tool for STAMP configuration in RAD planning was successfully developed. The proposed geometric cost-based framework achieved a dosimetric performance comparable to both expert-defined manual STAMP planning and the vendor-provided automatic collimator rotation approach, while eliminating the need for manual STAMP determination. These findings support the feasibility of the automated STAMP configuration as a practical strategy for enabling an efficient and planner-independent implementation of RAD planning.